IR: fix for loop over range with step

codeGenIntermediate/src/prog8/codegen/intermediate/IRCodeGen.kt

@@ -586,23 +586,22 @@ class IRCodeGen(
             else -> throw AssemblyError("invalid loopvar node type")
         }
         val loopvarDtIr = irType(loopvarDt)
-        val iterable = forLoop.iterable as PtRange
-        val step = iterable.step.number.toInt()
-        val rangeStart = (iterable.from as PtNumber).number.toInt()
-        val rangeEndExclusiveUntyped = (iterable.to as PtNumber).number.toInt() + step
-        if(step==0)
-            throw AssemblyError("step 0")
-        if(step>0 && rangeEndExclusiveUntyped<rangeStart || step<0 && rangeEndExclusiveUntyped>rangeStart)
+        val iterable = (forLoop.iterable as PtRange).toConstantIntegerRange()!!
+        if(iterable.isEmpty())
             throw AssemblyError("empty range")
+        if(iterable.step==0)
+            throw AssemblyError("step 0")
+        val rangeEndExclusiveUntyped = iterable.last + iterable.step
         val rangeEndExclusiveWrapped = if(loopvarDtIr==IRDataType.BYTE) rangeEndExclusiveUntyped and 255 else rangeEndExclusiveUntyped and 65535
         val result = mutableListOf<IRCodeChunkBase>()
         val chunk = IRCodeChunk(null, null)
-        chunk += IRInstruction(Opcode.LOAD, loopvarDtIr, reg1=indexReg, immediate = rangeStart)
+        chunk += IRInstruction(Opcode.LOAD, loopvarDtIr, reg1=indexReg, immediate = iterable.first)
         chunk += IRInstruction(Opcode.STOREM, loopvarDtIr, reg1=indexReg, labelSymbol=loopvarSymbol)
         result += chunk
         result += labelFirstChunk(translateNode(forLoop.statements), loopLabel)
-        val chunk2 = addConstMem(loopvarDtIr, null, loopvarSymbol, step)
+        val chunk2 = addConstMem(loopvarDtIr, null, loopvarSymbol, iterable.step)
         chunk2 += IRInstruction(Opcode.LOADM, loopvarDtIr, reg1 = indexReg, labelSymbol = loopvarSymbol)
+        chunk2 += IRInstruction(Opcode.XOR, loopvarDtIr, reg1 = 999, immediate = 111)
         chunk2 += IRInstruction(Opcode.BNE, loopvarDtIr, reg1 = indexReg, immediate = rangeEndExclusiveWrapped, labelSymbol = loopLabel)
         result += chunk2
         return result

compiler/test/vm/TestCompilerVirtual.kt

@@ -210,13 +210,13 @@ main {
         var result = compileText(target, true, src, writeAssembly = true)!!
         var virtfile = result.compilationOptions.outputDir.resolve(result.compilerAst.name + ".p8ir")
         VmRunner().runAndTestProgram(virtfile.readText()) { vm ->
-            vm.stepCount shouldBe 48
+            vm.stepCount shouldBe 59
         }
 
         result = compileText(target, false, src, writeAssembly = true)!!
         virtfile = result.compilationOptions.outputDir.resolve(result.compilerAst.name + ".p8ir")
         VmRunner().runAndTestProgram(virtfile.readText()) { vm ->
-            vm.stepCount shouldBe 48
+            vm.stepCount shouldBe 59
         }
     }
 

docs/source/todo.rst

@@ -1,9 +1,7 @@
 TODO
 ====
 
-- fix range expression problems in VM/IR (see test.p8)
-
-- investigate McCarthy evaluation again? this may also reduce code size perhaps for things like if a>4 or a<2 ....
+- [on branch:] investigate McCarthy evaluation again? this may also reduce code size perhaps for things like if a>4 or a<2 ....
 - IR: reduce the number of branch instructions such as BEQ, BEQR, etc (gradually), replace with CMP(I) + status branch instruction
 - IR: reduce amount of CMP/CMPI after instructions that set the status bits correctly (LOADs? INC? etc), but only after setting the status bits is verified!
 

examples/test.p8

@@ -10,6 +10,14 @@ main {
         ubyte xx
         byte bb
 
+        cx16.r1=0
+        for cx16.r0 in 0 to 10 {
+            cx16.r1++
+        }
+        txt.print_uw(cx16.r1)
+        txt.nl()
+        txt.nl()
+
         for xx in array {
             txt.print_ub(xx)
             txt.spc()